GitLab Patches Critical AI Gateway Sandbox Escape Allowing Arbitrary Command Execution
Emerging Technology Risk · 17 sources · Updated 2mo ago · First seen Oct 2, 2026

GitLab released AI Gateway versions 19.2.4, 19.3.2, and 19.4.1 to fix CVE-2026-90970, a critical improper neutralization vulnerability rated 9.9 on CVSS v3.1. Under certain conditions, an authenticated user with Duo Agent Platform access could submit a specially crafted flow configuration, escape the prompt template sandbox, and execute arbitrary commands on the AI Gateway. GitLab conducted targeted customer outreach before announcing the patches; the supplied notices do not report exploitation.
GitLab strongly recommends that affected Self-Hosted AI Gateway installations upgrade immediately to the applicable patched version. GitLab-hosted AI Gateways have already been patched, so GitLab.com, GitLab Dedicated, and GitLab Self-Managed customers using those hosted gateways need no action. The Canadian Centre for Cyber Security reinforced the update guidance in advisory AV26-994, identifying affected releases preceding the three fixed versions and urging administrators to review GitLab’s guidance and apply necessary updates.


